Global Agents Wanted! Join us to share market opportunities. Comprehensive training, technical support, and attractive incentives provided. Act now for a brilliant future! Contact email: partners@falenses.com
Machine vision, which has been developing for over 70 years, has reached a significant stage in its evolution. The early experiments in computer vision, the field that enables computers or machines to understand and interpret visual data to a certain extent, can be traced back to the 1950s. By the 1970s, preliminary commercial applications of machine vision had emerged. Today, machine vision has become an integral part of our daily lives, with applications ranging from facial recognition to translating road signs captured by cameras into our native language, or even selecting and copying text from images.
The early experiments in computer vision began in the 1950s, and by the 1970s, there were commercial applications that could distinguish between handwritten and printed text. In essence, computer vision allows computers or machines to interpret and analyze visual data, performing tasks such as observation, recognition, positioning, inspection, measurement, and decision-making that are traditionally done by the human eye. Machine vision systems can automatically acquire, analyze, and provide information from visual images to control machines or workflows.
The applications of machine vision extend far beyond facial recognition in photography. For instance, during a visit to Infineon's factory, it was observed that machine vision was used to inspect the yield issues of semiconductor products in certain processes, which is significantly more efficient than manual inspection. On a microscopic level, EDA and foundry factories also rely on machine vision to detect defects in chip manufacturing. Expanding further, when the term "Industry 4.0" became popular, it was clear that machine vision plays a crucial role in achieving industrial automation.
If we break down the workflow of computer vision into three main parts, it would roughly consist of image capture, image processing, and image analysis and understanding. Based on this, the cover story of this issue of "International Electronics Business" interviewed companies like SmartSens, Prophesee, Silvergate Microelectronics, and Imagination Technologies. SmartSens and Prophesee are mainly suppliers of image/vision sensors, focusing on image capture; Imagination is involved in the latter two areas—实际上Imagination has previously developed ISP (Image Signal Processor); Silvergate Microelectronics specializes in 3D vision, covering all three processes.
Although from an industry chain perspective, there are many other market participants, such as upstream light sources and lenses, midstream system integrators, and software and algorithm suppliers outside of hardware; however, we hope this article can outline the development potential of the current machine vision market. Let's discuss this field, which has a history of over 70 years, and see how it is faring now.
Machine vision (machine vision) and computer vision (computer vision) are terms that are often discussed in various contexts. There are differences in the definitions and distinctions between these two terms according to various sources. A report from iMedia Consulting previously suggested that machine vision is the engineering application of computer vision technology, "computer vision provides the theoretical and algorithmic foundation for image and scene analysis for machine vision, and machine vision provides sensor models, system construction, and implementation means for the realization of computer vision."
This statement seems to make sense and appears to be a different dimensional approach in different contexts. According to sources like Wikipedia, the term "machine vision" is more inclined towards the application of "computer vision" in the industrial field. Several experts we interviewed also shared similar views. For example, Mr. He Huogao, co-founder and vice president of Silvergate Microelectronics, mentioned: "Machine vision technology endows industrial equipment with the ability to 'see'," "Machine vision is a very important application field of computer vision technology, and computer vision is an important part of machine vision technology."
Rob Fisher, Product Director at Imagination Technologies, mentioned: "Machine vision can be considered a subset of computer vision, and computer vision encompasses a broader range of applications." Product Director Gilberto Rodriguez said: "The concepts of computer vision and machine vision are rapidly changing under the influence of machine learning." Prophesee experts mentioned that the boundaries between machine vision and computer vision are blurred, "In the same field, we often see them being used interchangeably. In our view, computer vision belongs to a broader field of visual technology, while machine vision is a subset of computer vision." "More specifically, we can regard machine vision as a set of task-oriented visual skills applied to specific applications (presence detection of objects, quality control, dimension measurement, automatic inspection, pass/fail decision...). Computer vision, on the other hand, is an interdisciplinary field that includes the most advanced visual perception and computation at the technical level."
In this article, we will no longer specifically distinguish between the terms computer vision and machine vision, and the discussion will be limited to the meaning of "machine vision" (for example, many industry reports and this article do not consider automotive ADAS systems as part of the machine vision category, even though they use many computer vision technologies; however, in some literature, these two terms can indeed be used interchangeably).
The second issue to clarify is the relationship between computer vision and AI. After all, we often see these two terms appear together. Understanding their relationship will also help us understand the development prospects of computer vision. What surprised us during our research is that almost all experts mentioned that computer vision is the application of AI in the visual world, and even said that computer vision is a subset of AI. According to our initial understanding, computer vision does indeed have some applications that use neural network technology, but this is not all of computer vision.
Later, we found that the general public's definition of "AI" is more broad, referring to the imitation of human behavior or other human characteristics and intelligence. Since machine vision is simulating the human eye and understanding in a specific field, it can naturally be classified as AI. However, in reality, the AI we often talk about now does not have such a broad definition.
Gilberto said: "Computer vision was originally used to describe algorithms written by humans and executed on general or specialized computing hardware. With the improvement of machine learning performance and the existence of efficient heterogeneous architectures, we can now obtain algorithms through training (AI training), without the need for humans to write code. This has changed our understanding of the concept of computer vision." He particularly emphasized that "with the increasing application of AI and machine vision technology, the use of traditional computer vision technology is decreasing." Here, the meaning of "AI" has actually narrowed down.
When we say that in 2012, AlexNet convolutional neural network (CNN) stood out in the ImageNet image recognition competition, and was also accelerated by GPU, this was a revolution in AI. And AlexNet was also considered at the time to be in the field of computer vision, the most far-reaching technological innovation. Then the concept of AI here has been narrowed down to deep learning (which is also the narrow definition of AI for many people). And the application of deep learning in computer vision is the major trend in the development of computer vision technology in the past two years. In this context, it is no longer appropriate to say that computer vision is a subset of AI; instead, it should be said that AI is driving the development of computer vision technology.
The rocket-like thrust of AI
AlexNet should be a representative of AI driving the development of machine vision technology; to the contemporary ResNet residual neural network, which often becomes a must-mention when we talk about AI and when AI chip companies release products. This actually shows that the development of contemporary computer vision is being pushed by AI. Just as Gilberto said, the proportion of traditional computer vision technology is significantly decreasing; or, from a programming perspective, those clear rules written based on human experience will become less and less competitive compared to AI technology.
"Especially in recent years, the boom of artificial intelligence has played a role in promoting the development of machine vision technology. AI has entered a new level, not just comparing computing power and indicators, but truly giving machines human characteristics and attributes. In the future, AI will play an increasingly important role in the field of machine vision and lead its future development direction." Mr. He Huogao said.
On the issue of applying deep learning to the field of machine vision, although we cannot provide exact numbers, it is evident from the companies we interviewed that the emphasis on AI is increasing. Imagination, for instance, currently promotes products other than GPUs, which are NNA (Neural Network Accelerator). Its GPU is also used for computer vision tasks, "such as 360° distortion removal, overlap, information display, etc." Gilberto said.
"Convolutional neural networks are very good at finding and classifying visual objects..." "NNA is one of our main computer vision processors. NNA is very suitable for deployment in high-efficiency machine learning computing, achieving the transition from traditional computing to machine learning." Of course, its RISC-V CPU is also a part of the pre-processing/ post-processing of machine learning tasks.
Silvergate Microelectronics focuses on 3D binocular stereo vision technology. It is worth mentioning that the company's self-developed NU4000 chip, Mr. He Huogao said: "NU4000 not only integrates deep learning engines, AI computing engines, general-purpose CPU cores, SLAM engines, but also enhances the expandability of application scenarios and greatly improves integration, truly becoming a single-chip solution SoC from deep perception to AI computing to system control. Following the integration of third-party DSP and CNN engines in NU4000, Silvergate will also integrate other AI processing capabilities into the next generation of chips in the future." He revealed that the future planned chip products will have a stronger main control CPU, better energy
FALenses Technology specializes in providing machine vision core hardware. You can go to the official website of FALenses Technology at https://www.falenses.com/ for more information.